I don’t think there’s much soldering involved. You need a hot air gun for mosfets.
Ya, FET definitely is the culprit in my two drivers.
I am glad it got sorted out and now I have over 32k lumens with two MT03 left and right. They got hot from cold really really fast. I plan to play with momentary mode side by side.
Thanks to you all who care, for all the help provided to me.
Yeah, I’ve got hot air. I just need to reflow quite a few more things before I take on my most expensive light.
That’s the craziest FET. I think you will notice at least 30% boost on brightness.
I didn’t practice, I was afraid of damaging FET with hot air, but it turned out working fine. Hot air temperature was 400c, I tried to complete it within 15s.
For modders, it is great light and really easy to open and repair or upgrade.
My FET is different from yours.
BSC009NE2LS5-I-ATMA1 vs. BSC009NE2LS5ATMA1.
I got the product number from Lexel but now I see that yours has lower RdsOn, higher current rating and lower price. Oh well. Anyway, it surely is a good FET. 
The key with hot air is distance. You can set an arbitrary temperature, on the high side, and control the heat by moving it closer or farther. Start off maybe 8 to 10 inches to heat up the FET and everything around it. After a minute or so have your tweezers ready and move your nozzle closer to the FET. Hopefully the solder will go liquid in about 10 seconds. Then lift it out of the way.
I usually remove the old solder with an iron and solder braid. Then I add flux and fresh leaded solder to the pads. Then add more flux and set the FET on top.
Then same thing, heat from a distance for a minute or so, then move in close looking for the FET to drop into position. Then I push it down with tweezers and remove the heat. Once cool, I clean any flux residue and it’s just like factory.
I’ve only done a few, but they came out perfect. If anyone has any more tips, I’m interested.
Thanks JasonWW. Will keep your hits at mind until I’m ready to do the repair.
You can also practice on junky circuit boards to get a feel for it.
Interested!
Your saying your interested in buying one and want the coupon code?
After all that has been said? Lol
I guess Freeme is still doing coupons for this light, I’m not sure. This thread was started in 2017. Just keep in mind there is a small chance you’ll need to replace the big FET on the driver.
I see. An extra “I”, yours is faster FET I guess.
Can you take picture on the FET when you got it, would like to see what’s the printing on the chip?
Now they can sleep in total darkness inside my camera bag that I have betrayed. 
Anyone knows is there a way to reprogram the low voltage protection through the firmware?
If it can’t be done through firmware, may I know which resistors to change and how is the calculation for resistor values?
You can adjust it in the firmware, although it can be tricky as you will need to use TK’s voltage hex file to figure out the right settings and then test it at a few voltages.
Is the voltage reading off? It is common for it to be off 0.1v, even 0.2v is not something I would worry about since there is a fair amount of margin for error on this.
Ya, one driver battery check is out by 0.2v. The LVP seems to kick on prematurely. So I thought if user can do manual offset via clicks on buttons.
Sadly no calibration is possible via the menu, only in the firmware code.
You can adjust the voltage divider resistors to correct the voltage but that is not always easy.
Although it may actually be working just fine. If you are running it in a high mode the voltage in the cells will sag lower, causing the LVP to kick in but when the load drops the voltage increases, which is why you get higher readings without the light being on.
It's something I want to add (i.e. voltage calibration), looked at it briefly, little involved but I'm getting, really really desperate to get this feature in for both voltage and temp. NarsilM uses voltage levels to the 100ths, not 10ths, so the data entry should be 2 digits - I want to make it work just like other config settings. In 100ths you can get it very accurate. In the firmware it's roughly 10 to 40 for 0.12V to 0.40V (usallly tighter from 0.20V to 0.36V or so) addition for taking the diode drop into account. I do this manually by checking batts at differing levels, again to 100ths precision.
In theory, I believe the diode I use supposed to be a ~0.24V voltage drop.
The variation from piece to piece is probably because of variances in the diodes and MCU's. I'm not really sure though if the bulk of variance if from driver design to driver design, or more from piece to piece. It's hard for me to tell because I generally reflow one driver at a time when I need it, and almost never use the same driver back to back.
To me, the ideal UI for this would be press&hold while in voltage display (blinking out) mode, the it goes into config, for data entry of 0.35V, blinks twice, then you click 3 times, stop, then it blinks twice again to confirm, then you click 5 times, and it blinks the final confirmation blinks - 4 quick ones. It should feel exactly like entry of other config settings.
Should work the same way for temperature calibration of entry of a 2 digit celsius temperature.
Tom, you might look at the Anduril code, I think TK has managed to get at least some of this implemented already. It would be cool if we could merge anduril and Narsil. They both have pros and cons and I think they could be blended together really well to get all the pros without the cons.
Yes, somewhat. Unfortunately I really don't like the turn OFF delay in Anduril - last I heard from her there's no fix for it, and not sure if she will fix it based on her response - it's not important to her, least not as important as it is to me. Actually I'm surprised no one else noticed or complained about it, or if they did, I didn't see it. It's something very basic with the most rudimentary action of the UI -- turning off the light, and it has an annoying delay... I don't get it.
If this delay is not fixed, I'm not gonna give up on NarsilM.
When I checked Anduril source code for voltage monitoring, I don't think she's using the full resolution that DEL developed and I brought into NarsilM, so she stuck with the old method with resolution in 10th's not 100's. Forgot the details, but DEL used higher resolution AtoD on the reading. He shared his code and I like it - dropped it right away into Narsil.